Hey,

Sasha here with the pleasure of delivering you some recent industry takeaways from April.

Unfortunately, it’s not all sunshine and rainbows in this edition of HIDDEN WISDOM #05.

Euripides once said “time will explain it all” in Aeolus back in 450 BC. Whatever he was referring to surely took longer than OpenAI to show its cracks. In roughly one quarter, the poster-child of the AI race posted its first real signs of strain.

OpenAI is now facing an existential crisis fueled by financial turmoil, legal troubles, and desperate pivots aside. Considering the incestuous commercial arrangements between LLM, Compute, and Hardware companies, the market reacted just about as you’d expect - quite strongly.

But are these surface cracks indicative of a bigger problem around AI?

Let’s dive in.

First we need to look at three things:

AI-related layoffs have steadily increased over the past six months.

The language is consistent in every story. Efficiency, optimization, automation.

The underlying story is less polished. Markets have been volatile, fiscal years have been weaker than expected, and AI is increasingly being used to justify restructuring that likely would have happened anyway.

The narrative around AI has outrun the product. A recent Princeton paper challenged the reliability of AI agents, pointing to inconsistent outputs and the continued need for human oversight. That doesn’t eliminate their value, but it does raise questions about how quickly they can fully replace human labor.

This leaves companies in an awkward position. Workforce decisions have already been made, while the actual efficiency gains are still playing out. And layered on top of this is a constraint that is much harder to ignore.

Energy.

Energy demand has risen sharply, driven by increased compute usage and broader geopolitical pressure.

At the same time, model developers continue to prioritize performance over efficiency. Anthropic’s upcoming Mythos model is expected to cost 5.5x more to run than its previous model. Industry commentary is starting to reflect this imbalance, with Nvidia’s Bryan Catanzaro noting that the cost of compute is now “far beyond” the cost of human employees in many cases.

The general sentiment toward AI is shifting:

Despite these signals, companies are not slowing down. If anything, they’re leaning in harder.

Energy is becoming a strategic resource, and the response has been aggressive:

We’re now seeing a world where AI expansion is directly tied to energy infrastructure buildout.

But experimentation hasn’t slowed. An AI-run business in San Francisco recently operated with minimal human oversight, making decisions across hiring and spending.

The current state of AI is not a clean exponential curve.

Token consumption has started to flatten against the increasing complexity of new model releases. Companies are moving slower than expected in replacing payroll, costs remain high relative to outcomes, and adoption is happening, but unevenly. None of this breaks the long-term trajectory, though.

Marc Andreessen’s idea that AI will eat application software still holds. AI is reducing the cost and complexity of building software, pushing the opportunity away from foundational models and toward the application layer.

The companies seeing traction are solving real business problems, moving quickly with smaller teams, and iterating faster than traditional development cycles allow. That’s also where capital is concentrating.

Arcanum Ventures believes the focus should be on AI-enabled applications that solve real business problems. In that sense, AI is becoming the next generation of software. And despite the broader concerns, areas like data authentication, privacy, and secure communication continue to present billion-dollar opportunities.

I’m hoping our insights help guide you through these uncertain markets.

We dove deeper into our 2026 thesis this month, covering industries like Prediction Markets,  Robotics, and other investment focus areas. And as always, we offered practical guidance for founders around storytelling.

Weekly recaps are on fire as listeners tune in to hear our takes on recent technology news and what the future holds for venture capital. 

The themes we’ve been seeing over the past few weeks: 

Carmelo took the stage to cover a range of topics and guide debate over some major happenings within the world of technology. 

As wearable technologies ramp up, it seems augmented reality infrastructure is taking the backseat. The industry seems to be focusing more and more on AI enablement as thought leaders claim that AGI is already here.

The reasoning? We can build billion-dollar apps….

This month leaned more negative than usual, but it reflects a broader adjustment. AI is not slowing down. It’s becoming more constrained, more expensive, and more grounded in reality.

For investors focused on fundamentals rather than momentum, that shift creates a clearer picture of where opportunities actually exist.

Thanks for sticking with me and happy May everybody.